dispatching-parallel-agents

Parallel Agent Dispatch (Dispatching Parallel Agents)

Skill Overview

When facing multiple independent failures or tasks, use this pattern to assign a dedicated agent to each independent problem domain, allowing them to work concurrently so multiple issues can be diagnosed and fixed within the same time frame.

Applicable Scenarios

1. Multiple test files failing simultaneously

When refactoring or code changes cause multiple test files to fail, and each file fails for different reasons (e.g., timeout issues, event structure errors, asynchronous execution problems), you can assign a separate agent to each test file to fix them in parallel.

2. Independent failures in multiple subsystems

When different modules of a system experience problems at the same time, and those modules do not share state or dependencies, you can assign an agent to each subsystem to investigate simultaneously, avoiding the time wasted by serial processing.

3. Concurrent handling of independent tasks

When facing two or more tasks that can be completed independently without sharing state (e.g., fixing bugs in different modules, reviewing different files, handling different types of errors), parallel scheduling lets multiple agents work at the same time.

Core Functions

1. Problem-domain identification and isolation

Help the user analyze whether multiple failures are independent and determine if parallel processing is appropriate. The core principle is: fixes in each problem domain should not affect others, and agents should not conflict with each other.

2. Dedicated agent task definition

Create focused task descriptions for each agent, including a clear scope, explicit objectives, constraints, and the expected output format, ensuring each agent can independently and efficiently complete its assigned task.

3. Parallel execution and result integration

Dispatch multiple agents to work concurrently, wait for all agents to return results, review each agent's summary, verify there are no conflicts between fixes, and then consolidate all changes.

Common Questions

When should parallel agent dispatch be used?

Consider parallel dispatch when you face three or more independent failures or tasks. The key criterion is: can these tasks be completed independently without knowledge of the others' contexts? If yes, parallel dispatch can significantly improve efficiency.

How do you determine whether tasks can be processed in parallel?

Check the following: (1) Does each task have clear boundaries and scope? (2) Will fixing one task affect others? (3) Will agents modify the same files or resources? If task boundaries are clear, tasks are independent, and there will be no resource conflicts, they are suitable for parallel processing.

How do you integrate results after parallel dispatch?

When all agents have completed their tasks, you should: (1) Read each agent's summary report to understand what changes were made; (2) Check for conflicts (e.g., multiple agents modified the same code); (3) Run the full test suite to verify that all fixes work together; (4) Make minor adjustments as necessary to integrate the changes.